[ETHANol]

This problem comes up over and over again. There are a lot of posts on it. Illumina seems to be clueless as how to address the issue (if they had given me that job I applied for, think about all the customer's headaches that could have been avoided). Bottom line is you really need to accurately determine the optimum number of PCR cycles you need to do for each and every sample. It is a trivial amount of time in the scheme of things and will reduce PCR bias. I outline an easy way to do it here:

http://ethanomics.wordpress.com/ngs-pcr-cycle-quantitation-protocol/

[monkey_SEQ]

Thanks for the reply Ethan. I will definitely implement your protocol in the future to test for optimum number of cycles of other samples. Unfortunately for these libraries I have no more tissue/ RNA left and I will only be able to get more in about 6 months or so. I wish Illumina would give better guidelines in their protocols so that newbies like me don’t just stupidly follow the protocol exactly only to realize afterwards that there are actually problems with it!!

So I size-selected a 300bp fragment (for what it was worth) and already sent it away for sequencing. I want to assemble a de novo transcriptome and then compare expression levels of different genes. In your experience, how much of an influence will an over-amplified library have on this experiment? Do you think I will be able to use the data at all?

[ETHANol]

As long as the sequencing goes well, the data should be fine. The biggest problem is the stress you went though wonder what went wrong and fearing that all that time you spent preparing your samples is down the drain. But it's not a big deal, which is probably why Illumna ignores it, but they should at least mention something in their kit instructions.

[monkey_SEQ]

Sigh, that's wonderful news And yes I think I might have a stomach ulcer by now!

[pbluescript]

Very nice. Could you please provide ladder size? Thanks!

Sure. It is an Invitrogen 1kb+ ladder, so the sizes visible are 100,200,300,400,500,650.

[pbluescript]

This problem comes up over and over again. There are a lot of posts on it. Illumina seems to be clueless as how to address the issue (if they had given me that job I applied for, think about all the customer's headaches that could have been avoided). Bottom line is you really need to accurately determine the optimum number of PCR cycles you need to do for each and every sample. It is a trivial amount of time in the scheme of things and will reduce PCR bias. I outline an easy way to do it here:
http://ethanomics.wordpress.com/ngs-...tion-protocol/

This looks even easier than the way I do it. Thanks!

[pmiguel]

This problem comes up over and over again. There are a lot of posts on it. Illumina seems to be clueless as how to address the issue (if they had given me that job I applied for, think about all the customer's headaches that could have been avoided). Bottom line is you really need to accurately determine the optimum number of PCR cycles you need to do for each and every sample. It is a trivial amount of time in the scheme of things and will reduce PCR bias. I outline an easy way to do it here:
http://ethanomics.wordpress.com/ngs-...tion-protocol/

Hi Ethan,
We are currently doing as low a number of cycles of enrichment as we think we can get away with. (3 cycles for some DNA TruSeq libraries). Then doing qPCR, then pooling based on the qPCR quantification. Then re-checking the 2nM lane pools with qPCR prior to clustering. (They are nearly always off.)

I was just wondering how your methodology was panning out, sample balance-wise?

Library quantification with qPCR seems to be replete with hidden mines that will blow up those so fool-hardy as to attempt to traverse its perilous fields without a map. And we are novices, really. Having only used qPCR for library titration, I am thinking there are some subtleties that we are not taking into account.

(Note these are all KAPA qPCR reactions -- SYBR Green and dilutions are always done with a buffer containing 0.1% TWEEN.)

One issue is that we often get high (as in "orders of magnitude") levels of variability in the initial sample concentrations (post enrichment PCR). The qPCR result-driven dilutions then rarely give accurate results when they are re-tested post-dilution. To mitigate this issue we did an entire extra round of qPCR after diluting all samples to 2nM -- prior to pooling.

Thinking about it, I guess that even though our standard dilution series (phiX lib from Illumina) gives a log-linear plot with an R-squared very close to 1.0, there is no guarantee that our samples would give a long-linear plot with the same slope. In fact, in cases were we do a couple of dilutions per sample, the resulting concentrations frequently disagree with each other.

I suspect nothing I wrote above would surprise a veteran qPCR researcher. So I was just wondering if you would like to go into what made you calibrate from the 50% amplification point, rather than the (traditional?) Ct?

--
Phillip

[Jon_Keats]

My lab is doing ETHANols pre-qPCR test on all of our CHIP and TruSeq RNA libraries and we are extremely happy with the results and I'd urge people to adopt the method if they are interested in reducing amplification bias and duplicate reads. In our RNAseq libraries we dropped the number of duplicate reads by 50% compared to the illumina protocol. In general 2ug inputs yield suggestions for 8 cycles of amplification.

In regards to the post amplifcation qPCR and dilution and re-testing the dilution we typically see our diluted pools re-QC within 10% of our anticipated value of 5nM. Most of the variation I attribute to small volumes of each library where pipet error, amount of DNA on the outside of the tip are causing fluctuations. We also do a large dilution series at both steps so we end up with 9 measurements. So right now I'm very happy with the qPCR quant at that phase.

Where we have issues with qPCR is it typically estimates molarity at 2 fold higher than bioanalyzer and if we target 600M/mm2 clusters on the HiSeq around 13pM from bioanalyzer seems to hit the mark but from qPCR 13pM is under-clustering.

[pmiguel]

What are you using for your standard with qPCR? We use the Illumina phiX libraries, but I have noticed that some of them are an average of 500 bp and some 350 bp. Specifications for the phiX version don't always seem to match the actual size. Get that wrong and you would be nearly 2x off. (Presuming you are using a SYBR green methodology for qPCR.)

--
Phillilp